Cluster connections are configured statically:

      <cluster-connections>
         <cluster-connection name="test-cluster">
            <connector-ref>artemis-master</connector-ref>
            <message-load-balancing>ON_DEMAND</message-load-balancing>
            <max-hops>0</max-hops>
            <static-connectors>
               <connector-ref>artemis-slave</connector-ref>
            </static-connectors>
         </cluster-connection>
      </cluster-connections>

Is there a way I can debug why this message appears in 2.30 and not in 2.28? 
I'm trying to avoid unforeseen consequences before upgrading production 
clusters to 2.30.

-- 
    Vilius

-----Original Message-----
From: Justin Bertram <jbert...@apache.org> 
Sent: Friday, September 1, 2023 6:45 PM
To: users@activemq.apache.org
Subject: Re: There is a possible split brain on nodeID XXXXX after upgrade to 
2.30

How is your cluster-connection configured? Are you using broadcast & discovery 
groups? If so, I've seen instances of this when stopping and starting the 
broker quickly. As far as I can tell the UDP multicast datagram somehow is 
delivered back to the broker that sent it originally.

As long as your backup broker isn't running while your primary broker is 
running then there's no issue.

FWIW, shared storage configurations are relatively immune to split brain due to 
the file locks enforced by the shared store.


Justin

On Fri, Sep 1, 2023 at 10:25 AM Vilius Šumskas <vilius.sums...@rivile.lt>
wrote:

> Hi,
>
> we have upgraded some of our environments from 2.28 to 2.30, others 
> from
> 2.24 to 2.30 and all upgraded instances now show this strange log 
> message during live node restart cycle:
>
> There is a possible split brain on nodeID 
> 165c6eec-0429-11ed-a12a-42010a961402. Topology update ignored
>
> Full restart cycle log https://p.defau.lt/?q4pc6mlUXzn66QY8zHVRjg
>
> The instances work normally and serve the clients through. I never saw 
> this happening on 2.28 or earlier versions, only 2.30 behave this way. 
> Is this something to worry about?
>
> All our instances are configured as HA live/backup clusters consisting 
> from 2 nodes. They share the data on NFS4.1 attached storage.
>
> --
>     Best Regards,
>     Vilius
>
>

Reply via email to